360 Secure SkillHub
360 安全技能中心

Lore

该 Skill 指导代理使用 Lore(MCP 研究知识库)的 ingest、search、research、get_source、retain 工具:摄取会议记录、访谈与外部内容,检索历史决策与项目背景,并在回答中附上引用来源。

mishkinf
mishkinf
热度 82.64
最近更新 2026-05-18 11:57:45
分类:知识依赖:MCP 服务应用场景:研究 / 笔记整理 / 摘要总结

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add lore
Metadata
name
lore
description
Search and ingest knowledge from Lore, a research repository with citations
version
1.0
user-invocable
false

Lore Knowledge Base

Lore is a research knowledge repository you have access to via MCP tools. It stores documents, meeting notes, interviews, and decisions with full citations — not just summaries, but the original content linked back to its source. Use it to ground your answers in evidence and to preserve important context from your conversations.

When to Ingest Content into Lore

Push content into Lore using the ingest tool whenever you encounter information worth preserving:

  • After conversations: When a user shares meeting notes, interview transcripts, or important documents, ingest them so they're searchable later.
  • External content: When you fetch content from Slack, Notion, GitHub, email, or other systems, ingest the relevant parts into Lore.
  • Decisions and context: When important decisions are made or context is shared that future conversations will need.

Always include:

  • source_url: The original URL (Slack permalink, Notion page URL, GitHub issue URL) for citation linking.
  • source_name: A human-readable label like "Slack #product-team" or "GitHub issue #42".
  • project: The project this content belongs to.

Ingestion is idempotent — calling ingest with the same content twice is safe and cheap (returns immediately with deduplicated: true).

When to Search Lore

Before answering questions about past decisions, user feedback, project history, or anything that might already be documented:

  1. Use search for quick lookups. Pick the right mode:

    • hybrid (default): Best for most queries
    • keyword: For exact terms, names, identifiers
    • semantic: For conceptual queries ("user frustrations", "pain points")
  2. Use research only when the question requires cross-referencing multiple sources or synthesizing findings. It costs 10x more than search — don't use it for simple lookups.

  3. Use get_source with include_content=true when you need the full original text of a specific document.

When to Retain Insights

Use retain (not ingest) for short, discrete pieces of knowledge:

  • Key decisions: "We chose X because Y"
  • Synthesized insights: "3/5 users mentioned Z as their top issue"
  • Requirements: "Must support SSO for enterprise"

Citation Best Practices

When presenting information from Lore, always cite your sources:

  • Reference the source title and date
  • Quote directly when possible
  • If a source_url is available, link to the original

Example Workflows

User asks about past decisions:

  1. search("authentication approach decisions", project: "my-app")
  2. Review results, get full source if needed: get_source(source_id, include_content: true)
  3. Present findings with citations

User shares meeting notes:

  1. ingest(content: "...", title: "Sprint Planning Jan 15", project: "my-app", source_type: "meeting", source_name: "Google Meet", participants: ["Alice", "Bob"])
  2. Confirm ingestion to user

User asks a broad research question:

  1. research(task: "What do users think about our onboarding flow?", project: "my-app")
  2. Present the synthesized findings with citations